Papers by Jiaqi Ye
AgentGym2: Benchmarking Large Language Model Agents in De-Idealized Real-World Environments (2026.acl-long)
Copied to clipboard
Zhiheng Xi, Dingwen Yang, Jiaqi Liu, Jixuan Huang, Honglin Guo, Baodai Huang, Tinggang Chen, Qi Zhang, Zhonghang Lu, Chenyu Liu, Jiajun Sun, Jiazheng Zhang, Dingwei Zhu, Xin Guo, Junzhe Wang, Zhihao Zhang, Yuming Yang, Junjie Ye, Minghe Gao, Dongrui Liu, Jiaming Ji, Guohao Li, Tao Gui, Qi Zhang, Xuanjing Huang
| Challenge: | Existing benchmarks evaluate agents in simplified, idealized settings, relying on pre-packaged tool interfaces, overlooking critical steps, and assume inputs are clean and fully specified. |
| Approach: | They propose a framework that evaluates language agents in simplified, idealized settings . they show that even SOTA systems like Gemini and GPT-5 struggle on AgentGym2 . |
| Outcome: | Experiments on 15 proprietary and open-source models show that even SOTA systems like Gemini and GPT-5 struggle on AgentGym2 . |
Automatic and Reliable Evaluation for Academic Caption-to-Figure Generation with LMMs (2026.acl-long)
Copied to clipboard
| Challenge: | Existing datasets for evaluating text-to-image generation focus mostly on real-life images, which poses challenges for assessing academic figure generation given real scientific captions. |
| Approach: | They propose a dataset that first provides a Holistic Evaluation for Academic caption-to-Figure Generation (HE4AFG) they collect real figure captions from 8 scientific domains and generate 3,900 evaluation samples . |
| Outcome: | The proposed model provides high-quality human ratings in terms of three aspects—scientific aesthetic (SA), topic relevance (TR), and attribute correctness (AC). |
Sentient Agent as a Judge: Evaluating Higher-Order Social Cognition in Large Language Models (2026.findings-acl)
Copied to clipboard
Bang Zhang, Ruotian Ma, Qingxuan Jiang, Peisong Wang, Jiaqi Chen, Zheng Xie, Xingyu Chen, Yue Wang, Fanghua Ye, Jian Li, Yifan Yang, Zhaopeng Tu, Xiaolong Li
| Challenge: | Large language models (LLMs) have evolved from statistical sequence predictors to sophisticated autonomous agents capable of reasoning, planning, and sustaining multi-turn conversa-tions. |
| Approach: | They propose a system that instantiates a "Sentient Agent" that simulates human-like emotional changes and inner thoughts to provide a more realistic evaluation of the model in multi-turn conversations. |
| Outcome: | The proposed framework measures the agent's higher-order social cognition in multi-turn conversations. |
Data Contamination Calibration for Black-box LLMs (2024.findings-acl)
Copied to clipboard
| Challenge: | Despite the rapid advancements of Large Language Models, the unchecked ultra-large-scale training sets introduce a series of potential risks like data contamination. |
| Approach: | They propose a method to detect contaminated training data and diminish the contamination effect by using a to-be-released dataset. |
| Outcome: | The proposed method outperforms existing methods by at least 4.5% on more 4 dataset formats, with more than 10 base LLMs. |
Shadow-Activated Backdoor Attacks on Multimodal Large Language Models (2025.findings-acl)
Copied to clipboard
Ziyi Yin, Muchao Ye, Yuanpu Cao, Jiaqi Wang, Aofei Chang, Han Liu, Jinghui Chen, Ting Wang, Fenglong Ma
| Challenge: | Existing backdoor attacks on Multimodal Large Language Models are less applicable to open-ended conversations with users. |
| Approach: | They propose a shadow-activated backdoor attack scenario where attackers inject malicious content into the responses of MLLMs when the responses explicitly relate to the shadowed object. |
| Outcome: | The proposed framework achieves the desired behaviors by constructing a poisoned dataset and implementing an attention-regularized tuning strategy. |
Bridging the Sensory Gap: Visual Injection for Taxonomy Completion (2026.acl-long)
Copied to clipboard
| Challenge: | Existing text-only methods suffer from a "Sensory Gap" in integrating new concepts into existing hierarchies. |
| Approach: | They propose a framework leveraging Visual Injection for Taxonomy Completion that maps synthesized images into intrinsic pseudo-tokens and decouples magnitude from selection to prevent visual signals from being drowned out. |
| Outcome: | Experiments on three datasets show that VITC achieves state-of-the-art performance . it delivers an average absolute gain of over 19% in Hit@1. |
WebSynthesis: World Model-Guided Monte Carlo Tree Search for Efficient WebAgent Trajectory Synthesis (2026.acl-long)
Copied to clipboard
| Challenge: | Recent advances in large language models have enabled increasingly capable web agents . however, training such agents at scale still relies on high-quality interaction trajectories that are difficult to obtain at scale. |
| Approach: | They propose a framework for scalable trajectory synthesis that simulates state transitions without network dependencies and integrates Monte Carlo Tree Search to enable reversible exploration over the simulated state space. |
| Outcome: | Experiments on WebArena, WebVoyager, and Mind2Web-Online show that agents trained exclusively on synthesized trajectories outperform those trained on real-world data. |
Towards Mitigating Modality Bias in Vision-Language Models for Temporal Action Localization (2026.acl-long)
Copied to clipboard
| Challenge: | Existing vision-language models overemphasize linguistic priors, leading to modality bias. |
| Approach: | They propose a vision-language aggregation framework that mitigates modality bias in TAL by preserving vision as the dominant signal while adaptively exploiting language only when beneficial. |
| Outcome: | Experiments on THUMOS14 show that the proposed model outperforms state-of-the-art models by up to 3.2% mAP. |